-
A Comprehensive Model for Human Factor Risk Assessment:
HFACS-FFT-ANN
Yu Liu1,a, Yang Liu1,b , Xiao-Xue Ma1,c* and Wei-Liang Qiao2,d 1
Public Administration and Humanities College, Dalian Maritime
University, Dalian, China
2Marine Engineering College, Dalian Maritime University, Dalian,
China [email protected], [email protected],
[email protected],
[email protected]
Keywords: HFACS, Fuzzy fault tree, Artificial neural network,
Fuzzy AHP.
Abstract. With the development of technology and reliability,
the human factors are becoming the most contributing casual for the
occurrence of accident. In the present study, a comprehensive human
factor analysis model, including Human Factors Analysis and
Classification System (HFACS), Fuzzy Fault Tree Analysis (F-FTA)
and Artificial Neural Network (ANN), is proposed to assess human
facors involved in accident or risk event. Under the framework of
proposed model, the FFT stemmed from HFACS is mapped into ANN based
on fuzzy theory, which may be beneficial for the prediction and
assessent of human factors.
1. Introduction Over decades, every efforts have been making to
hence safety navigation of ship at sea at national and
international level. However, a large number of marine accidents
still take place worldwide, resulting in immense loss of lives and
properties, also, the ocean environment is facing oil pollution
risk. The causal system leading to marine accident is characterized
by high level of complexity and uncertainty, however, it is
critical to implement the causal investigation for improving the
safety level of marine shipping. According to the statistics and
marine accident causal research, an overwhelming part of marine
accidents, about 75~96%, are attributable to human factors[1,2],
especially with the application of advanced technology in marine
shipping, this percentage may increase further.
The objective of the present study is to develop a comprehensive
model integrated by HFACS, Fault Tree Analysis (FTA), Fuzzy AHP, as
well as Artificial Neural Network (ANN), which is capable of
resolving the uncertainty of human factors involved in the
collisions at sea. The collisions at sea accident data in Chinese
waters is collected in this study to verify the proposed
methodology. In particular, the present study is characterized by
mapping the FT into ANN under the framework of HFACS, in which the
process digitalization and data mining are applied to establish
causal relationship by both machine learning and pattern
recognition. The proposed methodology in this study provides a
complete process (with steps and rules), from the development of FT
stemming from HFACS analysis, to the mapping of FT into ANN, which
may benefit the investigation of human factors involved in
collision accidents at sea.
2. Framework of the Proposed Model The overview of the
methodology proposed in this study is illustrated in figure a, in
which the integration of HFACS, FTA, H-AHP and ANN leads to the
formation of the comprehensive model to analysis human factors
involved in the collisions at sea. As shown in figure a, there are
totally 4 step in this study:
Step 1-Hierarchical Structure of the human factors is obtained
based on the HFACS framework. Step 2-modeling the Fault Tree
according to the hierarchical structure developed in step1. Step
3-huzzy APH is employed to calculate the failure probability of
Basic Events and Top Events
in the FT. Step 4-map the FT into ANN.
5th Annual International Conference on Management, Economics and
Social Development (ICMESD 2019)
Copyright © 2019, the Authors. Published by Atlantis Press. This
is an open access article under the CC BY-NC license
(http://creativecommons.org/licenses/by-nc/4.0/).
Advances in Economics, Business and Management Research, volume
87
148
-
Accident reports analysis
HFACS framework
Hierarchical structure
Unsafe acts
Preconceptions for unsafe acts
HFASC Unsafe supervision
Organizational factors
External factors
Fault tree analysis
Statistical analysis
Basic events
Intermediate events
Top events
Fault tree modeling
Causal classification
Causal frequency analysis
H-AHP
1-establish Fuzzy AHP2-Fuzzy aggregation analysis3-Convert
possibility to probability
Failure Probability of BEs and TEs
Basic events
Intermediate events
Logic Gates
Top events
Input nodes
Synaptic weights
Transfer function
Output node
Mapping process
Mapping FT into ANN
Figure 1. Overview of the proposed comprehensive model in the
present study
3. Application of the Proposed Model 3.1. Establishment of HFACS
Framework In the present study, HFACS-based risk analysis model is
developed to identify and classify the human factors involved in
the collisions. The proposed model consists of five layers and 22
categories, which is illustrated in fig.2. Obviously, this model is
established as a five levels framework, being similar to the
HFACS-Coll[3], HFACS-Grounding[4] and HFACS-MAM[5]. However, it
should be pointed out that the proposed HFACS model includes a
special layer, named as Hazard situation, which is regarded as a
supplement to the standard HFACS framework. Actually, the
introduction of “Hazard situation” is aimed at facilitating the
modelling process of Fault Tree, which will be discussed further in
the following section 3.2.
Advances in Economics, Business and Management Research, volume
87
149
-
Hazard situation
Level1
Level2
Level3
Level4
Level5
Collision at sea
Unsafe Acts Errors decision errors skill-based errors perceptual
errors Violations Routine violations exceptional
Preconditions for Unsafe Acts Situational factors physical
environment technology environment Cognition factors
mental/physiological states physical limitationPersonnel factors
fitness for duty communication/coordination
Unsafe Supervision Inadequate supervision planned
inappropriate
operators Failure to correct
known problem Supervisory violation
HFACS framework
Organizational influences Resource management Organizational
culture Organizational process
External factors Legislation gaps Administration oversights
Design flaws Social factors
Figure 2. The HFACS framework used in this study
3.2. Fault Tree Analysis Within the framework of FTA, the
failure probability of the system can be decomposed into different
kinds of failures and further into failures causes, until into the
elementary failure causes, which cannot be decomposed any more. The
typical FT is usually presented as a directed acyclic graph, which
consists of Top Event (TE), Basic Events (BEs), Intermediate Events
(IE) and gates.
TE
IEBE
BE BE
Collisions at sea
Hazard situations
Factors identified from L1-L5
Top Event
Intermediate Event
Basic Events
HFACS
Logic Gate
Logic Gate
Figure 3. Logic relationship between FT and HFACS
Advances in Economics, Business and Management Research, volume
87
150
-
In the present study, the elements, including TE, IEs and BEs,
involved in the FT can be transformed from the aforementioned HFACS
framework, and the transforming process is illustrated in fig.3.
The human factors causing to the occurrence of collisions at sea
identified from Level 1 to Level 5 under the framework of HFACS are
mapped into the structure of FT as the Basic Events. Meanwhile, the
“collisions at sea” in the HFACS framework corresponds to the Top
Events in the FT model. The hazard situations identified would be
transferred into Intermediate Events in FT, which is the reason why
the layer of hazard situation is introduced into the proposed HFACS
framework in this study.
3.3. Failure Probability Distribution of BEs and TE in FT Using
Fuzzy Theory Generally, there are three main approaches to
addressing the failure probability, statistical methods,
extrapolation, and expert judgment, respectively. In this study,
the expert judgment approach is chosen as a scientific consensus
technique to weight the identified human factors involved in the
collisions at sea and to rank expert capability. However, since
experts tend to express their opinions on each event based on their
individual knowledge, purpose and intellectual characteristics[6],
different analysis models have been established, such as fuzzy
priority relations, game theory, the max-min Delphi method and the
similarity aggregation method (SAM). It is hard to identify a
technique that is superior to the others for aggregating expert
opinions; however, it is widely accepted that ambiguous expression
from experts is extremely common. Thus, the integration of fuzzy
set theory and the Analytic Hierarchy Process (AHP) is frequently
utilized to aggregate experts’ ambiguity.
3.4. Mapping FT into ANN With the development of computational
tools, artificial neural networks (ANN) are created gradually, and
have been widely applied as computational techniques for modeling
and forecasting in various fields[7]. As is shown in fig.4, the
training and testing of network are two main steps based on the
establishment of ANN architecture. For the training of the network,
the back-propagation algorithm is widely applied for tuning
suitable weights of each neuron. The two-step iterations involved
in the back-propagation algorithm: the first forward step
calculates the results; the second backward step is aimed at error
computations and weights updating. The iterations would
continuously run until the calculated error satisfies the
pre-defined goal tolerance requirement[8]. Among the lists of
methods utilized to minimizing the overall error, the Mean Squared
Error (MSE) is preferred, and widely used.
nx θ 1x 2x 3x
nw2w 3w1w 0w
f
ijw output
Summing function
Transfer function
TE
IEBE
BE BEBasic Events
Figure 4. Mapping FT into ANN
Advances in Economics, Business and Management Research, volume
87
151
-
4. Summary In the present study, a comprehensive model is
proposed combining fuzzy set theory, HFACS, FTA, AHP, and ANN.
Focusing on human factors, the model aims to identify, characterize
and rank the human factors involved in accidents or risk events
from a causation perspective. The proposed model can effectively
handle the uncertainty and intuitive opinions of experts regarding
sand carrier accident analysis.
5. Acknowledgement The authors gratefully acknowledge the
financial support provided by the Fundamental Research Funds for
the Central Universities (grand No. 3132019190).
6. References [1] Ugurlu, Application of fuzzy extended AHP
methodology for selection of ideal ship for oceangoing watchkeeping
officers, Int. J. Ind. Ergonom, 2015, 47(5)132-140.
[2] D. Pennie, N. Brook-Carter, W. Gibson, Human factors
guidance for maintenance. In: Human factors in ship design, safety
and operation conference. The Royal Institution of Naval
Architects, March, London, UK, 2007.
[3] C. Chauvin, S. Lardjane, et al., Human and organisational
factors in maritime accidents- Analysis of collisions at sea using
the HFACS. Accid. Anal. Prev. 2013, 59, 26–37.
[4] S.T. Chen, A. Wall, et al., A human and organisational
factors (hofs) analysis method for marine casualties using
hfacs-maritime accidents (hfacs-ma). Saf. Sci. 2013, 60 (12),
105–114.
[5] Mazaheri, J. Montewka, et al., Usability of accident and
incident reports for evidence-based risk modeling – a case study on
ship grounding reports. Saf. Sci. 2015, 76, 202–214.
[6] D.N. Ford, J. Sterman, Dynamic modeling of product
development processes. Syst. Dyn. Rev. 1998, 14 (1), 31–68.
[7] N. Chitsazan, A. Nadiri, et al., Prediction and structural
uncertainty analyses of artificial neural networks using
hierarchical Bayesian model averaging. J. Hydrol. 2015, 528,
52–62.
[8] I.A. Basheer, M. Hajmeer, Artificial neural networks:
fundamentals, computing, design, and application. J. Microbiol.
Methods 2000, 43 (1), 3–31.
Advances in Economics, Business and Management Research, volume
87
152